Will the Real FCoE Standards Stand Up ? No, they can’t.

It’s pretty well known that I am not a big believer in FibreChannel, or even worse, FibreChannel over Ethernet.

But J Michel Metz ((From his blog: “Most recently J has used his skills as a Solutions Architect/Marketing Manager to espouse and promote Fibre Channel over Ethernet (FCoE), its promise and future in the data center, as well as promote over 20 joint QLogic/HP products with internatonal routes-to-market as a focus.” )) has been recruited by Cisco to go out into the community and “evangelize” the FCoE protocol. So, we have a blog post on Cisco’s Data Center blog that claims that FCoE standards are all packed up and ready to go.

Here is the link :
Will The Real FCoE Standards Please Stand Up

Read that. Yep, feels like misdirection doesn’t it. Like a magic trick: “No, don’t look at my right hand, look at MY LEFT HAND.”

“Oh, NO they’re not”

While it’s true that the FibreChannel standards that make up the least significant part of the FibreChannel over Ethernet protocol have been complete for a long time, you can’t claim a moral leadership because of that and that, by extension, all the other standards are under control. Cisco has been claiming that DCB Ethernet standards will be ready since the middle of 2009. Now it’s late 2010 and still no sign of completion.

So, we have a car and no wheels. Or gearbox. Just because Cisco chose to put out a range of products that use a version of FCoE doesn’t validate the technology. Neither does the half a billion dollars that Cisco has spent so far in technology and marketing. It’s certainly enough to create some buzz and force partners to do something about it.

The EtherealMind View

FCoE is clearly struggling. Cisco is the only vendor throwing full support into it. I believe that other vendors feel pressured by Cisco to add support. Whether indirectly, because of Cisco’s dominant market position, or directly as Cisco uses partnership agreements to lock in technology adoption, really doesn’t matter. I say this because only Cisco’s partners such as EMC and NetApp are doing anything. CNAs are coming from two or three companies, but almost no one else. HP, IBM, Brocade ? Yeah, it’s coming they say. I’ve been saying similar things since April 2009 The Case Against FibreChannel.

In the meantime, the momentum behind NFS and iSCSI as viable storage networking tools is growing. Next year, SAN booting over iSCSI is expected to be widely available, as will CNAs with high speed iSCSI and NFS performance. Check Microsoft and Intel Push One Million iSCSI IOPS There are reasons why Microsoft has not released a native FCoE client for Windows Server (because creating a lossless network driver is difficult for Microsoft to achieve).

Don’t let Cisco run out the astroturf and hide the fact that DCB standards are not here, and won’t be here anytime soon. Just because Cisco wants you to use their proprietary, pre-standard technology today so they can some ‘leadership’.

And when they do come, you will need to forklift all of your Cisco kit out of your datacenter to get the new features. That’s the prize for Cisco, and that’s what drives this ‘marketing exercise’. FCoE has a place in just a few data centres of the future, but for the vast majority, don’t waste your money on it. Get proven, reliable storage networking using NFS and iSCSI and use your existing network equipment. Plan for migration as your networks grows and DCB will provide you the tools to scale IP storage protocols.

Don’t listen to the paid marketing message from Cisco. Make your own decision.

  • http://charleshood.net Charles Hood

    Interesting perspective Greg, but one small correction: Brocade has indeed been shipping CNAs since Spring 2009.

  • http://www.twitter.com/jmichelmetz J Michel Metz

    Greg,

    First of all, I’m particularly honored that you’ve dedicated an entire post to, um, me. Gawrsh, look at me, I’m blushing!

    I’m also rather chuffed that you’ve not only retweeted the link but posted it so prominently in your blog. I definitely appreciate the promotion!

    It seems, though, that there are a couple of things you may have missed in the process, however, and as a result the bait-and-switch appears to be more on your side than mine.

    For one thing, my post was designed to address some of the inaccuracies being reported in the trade press about standards. Quite frankly, that’s it. It’s not clear how that was “misdirection.” After all, I stated what I was going to do, and did it. If you can be a little more specific as to what you think I claimed I was going to write about but didn’t, let me know; I certainly have no intention to dissemble.

    So, you imply that standards are unimportant, but in reality they’re VERY important to a lot of customers. Apparently they’re important enough for them to be brought up in nearly every article about FCoE (whether you care about FCoE or not, Greg, isn’t the question). As a result, if you want to have an honest discussion about FCoE and it’s value – such that it may be – it’s important to at least be accurate.

    So, for those who wish to have discussions based upon whether FCoE or Converged I/O is appropriate for them, my contention is that it’s more useful to have a decent understanding of what is, or is not, accurate.

    Now, as far as your claim that the DCB standards aren’t finished, this is difficult to give a blanket answer to. Why? because DCB is a group of standards to deal with solving Ethernet problems, not just for FCoE. However, all the standards that directly relate to FCoE are technically stable:

    802.1Qbb (PFC) was completed in July 2010 and is now awaiting publication
    802.1Qaz (ETS & DCBX) was also completed in July 2010 and has moved out of development phase and into approval phase (in other words, all the technical changes are complete, it’s stable, and consensus has been achieved).

    That’s it. Any company who wants to wait until publication to *begin* working on FCoE will be *years* behind those who started when the standards have stabilized.

    You also seem to be under the impression that Cisco is going at it alone. I’m sure that Brocade, QLogic, Emulex, HP, IBM, EMC, NetApp, Intel, and a whole host of other companies would be very surprised to find out they haven’t actually made any progress. IBM has had CNAs since mid-2009, for instance, so it’s more than just “it’s coming.” HP announced a QLogic-powered FCoE module for its BladeSystem in June, 2010.

    Finally, you use the term “astroturf” in a way that doesn’t make sense. Astroturfing is, in the vernacular, when someone attempts to falsely initiate a “grassroots” movement on neutral ground. As you point out, I’ve been completely open about my past employment as well as my current role. No falseness either implied nor enacted. Also, as it was on Cisco’s own website, it’s difficult for the claim of “astroturfing” to stick; that’s not exactly neutral ground, is it?

    So, in the immortal words of Inigo Montoya, “You keep using that word. I don’t think it means what you think it means.”

    Look, I’m not saying that you don’t bring up good points as to the value of FCoE/iSCSI/NAS/etc. But my post wasn’t to talk about value propositions, nor was it to talk about FCoE implementations or adoption. All my post was intended to do was address some of the inaccuracies about standards that were being reported. I did that, and it was you who decided to use that as a launching pad to talk about what you wanted to talk about.

    Not that there’s anything wrong with that; I mean you can definitely use whatever catalyst you like to launch your own ideas. But you can’t really get upset with me for not writing what you want to talk about and then accuse me of misdirection, now can you? Well, you can, obviously, but it doesn’t hold up under scrutiny.

    Nevertheless, I do appreciate the links! I also enjoy the debate. Perhaps next time you’ll invite me to it personally.

    J

  • http://etherealmind.com Greg Ferro

    This is the same debate that I have been having for more than 18 months. FCoE still has very little market momentum and most customers are not confident in the technology. All the vendors you list are slowly delivering something, but I would maintain it’s a ‘just in case’ strategy because Cisco is the 800 pound gorilla, and not because the technology is obviously or even inherently valuable.

    The very fact that Cisco has recruited you to be a “evangelist” speaks volumes about the lack of success. It must be in dire straits to waste money like that instead of focussing on sales and new products.

    BTW, the Ethernet standards were originally predicted by Nuova Systems to be complete in mid-2009. Regardless of what sophistry you choose, most customers are going to careful about pre-standards technology. The few Cisco-centric customers that have purchased Top of Rack solutions are not signs of mass technology adoption and deployment in my view. For comparison, consider the Ethernet vs Token Ring uptake.

    I understand exactly what astroturf implies. I perceive that you claiming that FCoE standards are complete is an attempt to coverup the fact that DCB are not. And FCoE does not work without DCB. To attempt to diffuse the damaging fear of Cisco moving into a locked-in technology such as FabricPath you are using indirection to claim that it’s all “OK”, and it isn’t.

    PS. You shouldn’t feel honored, I’ve engaged in this debate with many people, you are just the latest who hasn’t yet seen the light.

  • http://www.twitter.com/jmichelmetz J Michel Metz

    If there is an element of discomfort about the technology, it is because there is very little understanding about it. The sheer amount of misinformation – in this case about the nature and status of technological standards – does nothing to ameliorate this problem. My post was an attempt to provide some clarification on where we are in the standards process so that those who are interested in finding out more information can do so with some facts as starting points, rather than unfounded assertions.

    The sad truth is that because of the nature of my position I am unable to comment on adoption rates for FCoE. I am not permitted to name names or numbers. If you want to beat me up on that fact, well, there’s nothing I can do so wail away – nuttin’ I can do about that. But I can say that you’re simply wrong about your assertions in take-up. Given that you have equally about as much ability to back up your claim as I do, that argument is pretty much a wash.

    Finally, I’m not sure how to respond to your astroturfing comment. It’s simply wrong. FCoE standards *are* complete. They’ve been technically stable since October 2008. DCB standards (at least the ones that pertain to FCoE) are technically stable as well, and in at least one case moving towards publication. Whether or not Nuova predicted their completion by 2009 is irrelevant.

    FWIW, FCoE *can* work without DCB, just the way we choose to implement it requires one element of DCB: PFC. In five or ten years, we may choose a different way to provide lossless traffic that doesn’t involve PFC – who knows? But the point is that both FCoE *and* PFC are complete. If we’re talking about FCoE and standards (which I was), TRILL/FabricPath don’t even enter into the conversation. Talk about a red herring! You accuse me of right hand/left hand shennanigans, and then you pull that doozy.

    Again, I stated what my purpose of my blog was (FCoE standards), and I addressed it. So far you have not been able to counter either the purpose nor the delivery of what I’ve *actually* written, instead focusing on other arguments that were never part of the scope in the first place: “J spends his entire post talking about standards, and not talking about what I have been talking about. Therefore he’s *misleading* you!!”

    Sorry, but it doesn’t hold water just because you say it does.

  • Time to Wake Up

    I knew from the first line not to trust what you said with regards to this: “I am not a big believer in FibreChannel”. Right there is where the problem is. Having a discussion, let alone argument, with someone about a topic they don’t believe in right from the start is just a waste of time. Let’s argue evolution with a devout Christian, doesn’t matter what you say, they will never agree.

    Fact of the matter is this. Data Centers that have large data requirements are using FC, period end of story. Facts are facts, accept it. As server density, blades, VM etc have accelerated people are looking for technologies that simplify their lives, which is something that FCOE does. While iSCSI isn’t terrible, it has never had the throughput of FC. FCOE at 10Gig has similar throughput to 4gFC, while iSCSI doesn’t.

    Was FCOE created by Cisco, yes. Are the standards completed, for FCOE yes. Will FCOE work in today’s DCs as a replacement for iSCSI and FC, yes. Will a 2 TB Oracle DB be moved to FCOE or iSCSI or NFS, definitely NOT (yet).

    FCOE, iSCSI, NFS, and FC are all tools. And like with any tool it has a purpose. Use the correct one for the correct situation and you will be very pleased. Killing a Fly with a sledgehammer will work, but its wrong. But try to kill a NYC RAT with a fly swatter just won’t work. I will let you sort out which tech is which tool.

    • http://etherealmind.com Greg Ferro

      You make the assumption that FibreChannel is the only protocol. Storage Arrays are not limited to using FibreChannel and most vendors support multiple protocols. Therefore a discussion around which protocol is most effective for the entire industry is valid.

      Second, by declaring my position, I make it clear where I stand. Regular readers most likely do, new readers do not.

      My overall position is that FCoE is a return to Token Ring or FDDI networking. Good technical reasons drive the idea, but the implementation is ultimately flawed, costly and bad for our image as Networking Professionals.

      There are many customers moving to iSCSI and NFS storage networks exclusively. But no one hears about them because they don’t have a marketing budget like Cisco. And that’s what J Metz is doing, using the market position and wealth to create a distorted view of the industry.

      • Time to Wake Up

        I am assuming nothing, just going based on facts. There is a tremendous market for FC right now, has been for years. Are iSCSI and NFS doing well, sure, but I would never allow NFS to flow through a firewall (what a nightmare).

        FCOE is costly? Really how so? Less cards, less cables but similar throughput to the 4gigFC solutions.

        Additionally if FCOE is proprietary then how come QLOGIC, Intel, Emulex, NetApps, EMC, 3Par and the like are using this protocol. Its not like EIGRP, it is a standard, a developing one, yes but a standard none the less.

        There are times I really enjoy your posts but there are times where I wish there would be more input from others to help flesh out a pro/con concept.

        • http://etherealmind.com Greg Ferro

          NFS works OK through a firewall if you must. At least you have the choice to do that and there are marginal use cases for it.

          FCoE requires an enforced forklift of all networking equipment since existing assets in the data path must be replaced.

          As a I wrote, the reason other vendors are supporting FCoE is because Cisco says that’s the way to go. Further, Cisco has splashed down half a billion dollars to show they are serious and so everyone follows along. There isn’t any discussion of whether its a good choice, as a Cisco partner you are forced to join in or miss out on sales opportunities.

          It seems like I’m one of the few either for or against. Most people are just blindly doing what the man says. That makes me sad.

          • Time to Wake Up

            ‘Enforced forklift of all networking equipment because everything must be replaced’

            Really? How.

            So if you already have a FC backbone in your environment you can add an additional device to the mix that supports FCOE (either blade or switch). Devices that are talking FC only can talk to the FCOE devices and the other way around. Why is that so hard? Even in a greenfield environment you can have it all and choose the right tool for the job.

            Your comment about Cisco beating everyone into it overstates the issue completely. For years Cisco said FAT Access Points for Wireless were the solution, and Airespace, Aruba, Trapeze and others disagreed and kept on moving forward, until Cisco realized it and bought one of them. If EMC/NetApp/Brocade/3Par/QLogic/Emulex disagreed they would let Cisco flounder with it and lose money as they gained.

            FCOE is a completely viable solution, have you implemented it? Since you state your hatred for FC my guess is a NO. I can truly say that I would never allow NFS through a firewall, because I have done it and its a nightmare.

          • http://etherealmind.com Greg Ferro

            I’ve designed and worked with FC and FCoE to a limited extent, and it reminds me of the FDDI switching backbones I built using DEC equipment in 1998. Expensive adapters, arbitrary rules, pointless implementation, limited choices and on and on. It’s a fine technology, but that doesn’t make it a good choice, and the market has already agreed that FC is over. FCoE is a transition manoeuvre at best to destabilise FC vendors that has gotten out of hand.

            While I’m not a fan of ethernet for this technology, it’s inevitable. Customers and resellers are not wise enough to make better choices. If you are going to use Ethernet, then don’t use FibreChannel because the layering of lossless technology over a lossy network is going to be painful. A tragic case of turd polish.

            I standby the point of Cisco forcing or driving it into the market. If it wasn’t Cisco, no one else would spend the sort of dollars to develop the technology.

          • http://etherealmind.com Greg Ferro

            Readers should note that this commenter is an employee of Cisco and weigh the point of view accordingly. Note also that they failed to disclose the affiliation.

          • Time to Wake Up

            Fun discussion BTW…..

            Cost is an interesting one though. We just bought a pair of CNAs ($2000), but if I was to buy a pair of 4gFC HBAs ($1800) and then the extra FC ports. There is a savings here, and that does factor into it.

          • Time to Wake Up

            Sorry yes I do work for Cisco but the facts remain the facts. Cost, simplicity and functionality. We should be able to have it all. I don’t believe I stated anything that smacks of CISCO KOOLAID, if I did then I apologize.

            We all need to look at these technologies for what they are and what they can offer us. Sometimes its about Cost, performance, manageability, reliability, scale, and many other factors. I think all the technologies we discussed are viable but for specific purposes.

            I pose my earlier question again: Would you recommend NFS/iSCSI to a company running a 2TB Oracle cluster? I don’t know if I would recommend FCOE yet, but FC is a perfect solution for that instance.

  • http://wikibon.org/blog Stu Miniman

    Greg,
    We know how biased you are against all things FC. The question that I always have for you is that for the large enterprise customers, what solution do you have that can match the scalability, manageability and reliability of FC for thousands of nodes in a single network? iSCSI is NOT the answer – it is great for a couple of hundred nodes. NFS is starting to grow with scale-out architectures, but moving from FC to NFS is complicated. Moving FC to FCoE gets everyone on Ethernet and down the road we can talk about what the “end-stage” solution will be. You are correct that DCB requires new hardware and changes, but that is going to be much easier to migrate FC > FCoE than any other Ethernet option. For the customers that would use iSCSI or NFS, FCoE isn’t likely an option – so no need to be so dogmatic on the arguments.

    • http://etherealmind.com Greg Ferro

      That’s why I state by position up front.

      Solution: You should ask the major global consulting firm who recently built an outsourcing data centre of fifteen hundred servers with iSCSI or NFS for every host and now scaling to thousands more. FC didn’t add up for them in terms of cost or features.

      There is nothing complicated about iSCSI or NFS, they ahve been around for years and are quite well understood. I believe it’s FibreChannel that needs training, and specialist knowledge, special equipment, with massive cost overheads as it attempts to build a lossless network.

      Your assumptions come from the bigoted industry position that only fibrechannel can make storage work, which is patently untrue. Customers need to consider iSCSI or NFS as the FIRST choice. FCoE means you have failed to comprehend the challenge and you have chosen the safe, expensive and vendor profitable option.

  • Chris Fricke

    I managed a fibrechannel SAN – granted not a very big one, only two racks worth of pieces parts – for many years and when it came time for performance upgrade we indeed forklifted everything… to iSCSI. And haven’t looked back (what’s not to like about double the capacity, better performance, and 25% of the cost and physical space). That was in 2004 and some called the move crazy but the reality is we could not have sustained our rate of growth in FC dollars.

    FC absolutely has a place in many environments. There are just as many datacenters, though, that should really take a hard look at their environment and reevaluate the problems they are trying to solve. For most workloads FC/FCoE just doesn’t make sense anymore.

  • Pingback: FCoE Frustration? Me too. « J Metz's Blog()